Eaton and Palisades Fires: A Satellite & Social Vulnerability Analysis

Github Repository: https://github.com/jwonyk/ep-wildfire-analysis

About

Purpose

In early 2025, two major Southern California wildfires, the Eaton and the Palisades Fire, burned across different parts of Los Angeles County. To better understand the impacts of these events, this project combines Landsat 8 satellite imagery with census-tract-level community indicators from the 2024 Environmental Justice Index (EJI). By integrating remote sensing and social vulnerability data, this analysis highlights both the visible physical burn patterns from space and the community characteristics of affected populations.

Highlights

• Restore missing CRS from a Landsat NetCDF file
• Create true and false-color images to visualize vegetation, burned areas, and the landscape condition
• Overlay false-color imagery with the fire perimeter to illustrate aftermath of wildfire
• Examined key vulnerability indicator: age
• Compared the vulnerability profiles of communities affected by the Palisades and Eaton fires

About the data

Fire Perimeter Dataset

The fire perimeter dataset contains dissolved burn boundaries for the Eaton and Palisades fires. The original data comes from NIFC FIRIS (Fire Integrated Real-time Intelligence System). The data provides daily fire growth during the event of wildfire. The dataset is accessed through Los Angeles GeoHub and it is also accesscible from NIFC FIRIS public ArcGIS service.

Landsat NetCDF Dataset

The Landsat used in this analysis is a Landsat 8 Collection Level-2 Surface Reflectance product accessed through the Microsoft Planetary Computer Data Catalogue. The dataset includes multiple spectral bands stored in a NetCDF structure, CRS information, coordinate dimensions, and metadata. The spectral bands will assist to create true and false-color composite images for post-fire assessment.

Environmental Justice Index Community Vulnerability Analysis

This analysis uses the 2024 Environmental Justice Index (EJI) to identify and compare census tracts affected by the Eaton and Palisades fires, using spatial joins and clipping to the fire boundaries. After aligning datasets to a shared CRS, the intersecting tracts are examined using four key vulnerability indicators, such as age, vehicle access, poverty, and overall EJI score, to highlight differences in social and demographic vulnerability between the two burn areas.

Part 1 - Satellite Imagery Analysis

Setup for the Analysis

# Import packages for this project
import os
import numpy as np
import pandas as pd
import geopandas as gpd
import xarray as xr
import rioxarray
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches


# Set pandas to display all columns
pd.set_option("display.max_columns", None)

# Load fire perimeter datasets
eaton = gpd.read_file(os.path.join('data',
                                      'fire_perimeters',
                                      'Eaton_Perimeter_20250121.geojson'))

palisades = gpd.read_file(os.path.join('data',
                                      'fire_perimeters',
                                      'Palisades_Perimeter_20250121.geojson'))

# Import NetCDF dataset
landsat = xr.open_dataset("data/landsat8-2025-02-23-palisades-eaton.nc")

Restore Geospatial Information (CRS Recovery)

The NetCDF file contains projection information inside the spatial_ref variable. rio.crs initially shows that no CRS is assigned. I will extract the WKT projection from spatial_ref.crs_wkt and write it back to the dataset using rio.write_crs().

# extract CRS from the spatial_ref variable
crs_wkt = landsat.spatial_ref.crs_wkt

# write CRS into rioxarray metadata
landsat = landsat.rio.write_crs(crs_wkt)

True-Color Composite

I will select the red, green, and blue bands and converted them into a 3-band array. The initial plot produces warnings due to bright cloud outliers, so I will use the robust = True parameter to ignore extreme values. At the end, I will identify and replace NaN values to create a clean final RGB image.

rgb = landsat[['red', 'green', 'blue']].fillna(0)

rgb_plot = (rgb.to_array().plot.imshow(
    robust = True, 
    figsize = (7, 7), 
    add_colorbar = False))

plt.title("True Color Composite (Landsat 8, Feb 23 2025)", fontsize = 14)
plt.axis("off")
plt.tight_layout()

False-Color Composite

False color composites help highlight vegetation health and burn severity. I will plot a SWIR2, NIR, Red composite without creating new variables.

(landsat[['swir22', 'nir08', 'red']].fillna(0)
                                  .to_array()
                                  .plot.imshow(
                                    robust = True, 
                                    figsize = (7,7),
                                    add_colorbar = False))

plt.title("False Color Composite (Landsat 8, Feb 23 2025)", fontsize = 14)
plt.axis("off")
plt.tight_layout()

Overlaying Fire perimeters

# Create a figure and axes for plotting
fig, ax = plt.subplots(figsize = (10, 10))

# ---------------------------------------------------------------
# BASE LAYER
# ---------------------------------------------------------------

# Plot the false color landsat composite (SWIR2, NIR, Red)
(landsat[['swir22', 'nir08', 'red']].fillna(0)
                                    .to_array()
                                    .plot.imshow(ax = ax,
                                    add_colorbar = False,
                                    robust = True))

# ---------------------------------------------------------------
# FIRE PERIMETER OVERLAYS
# ---------------------------------------------------------------

# Plot Eaton Fire perimeter using the raster's CRS
eaton.to_crs(landsat.rio.crs).plot(ax = ax,
                                   edgecolor = 'yellow',
                                   facecolor = 'none',
                                   linewidth = 1.5)

# Plot Palisades Fire perimeter using the raster's CRS
palisades.to_crs(landsat.rio.crs).plot(ax = ax,
                                   edgecolor = 'red',
                                   facecolor = 'none',
                                   linewidth = 1.5)

# ---------------------------------------------------------------
# CUSTOM LEGEND PATCHES
# ---------------------------------------------------------------

# Create a custom label bubble for Eaton
eaton_patch = mpatches.Patch(facecolor = 'none',
                             edgecolor = 'yellow',
                             linewidth = 2.5,
                             label = 'Eaton Fire')

# Create a custom label bubble for Palisades
palisades_patch = mpatches.Patch(facecolor = 'none',
                             edgecolor = 'red',
                             linewidth = 2.5,
                             label = 'Palisades Fire')

# Display a legend identifying each fire perimeter
ax.legend(handles = [eaton_patch, palisades_patch],
          loc = 'upper left')

# ---------------------------------------------------------------
# ON-MAP TEXT LABELS
# ---------------------------------------------------------------

# Eaton label (top right)
ax.text(0.75, 0.80,
        'Eaton Fire',
        transform = ax.transAxes,
        fontsize = 12,
        color = 'yellow',
        fontweight = 'bold',
        bbox = dict(boxstyle = 'round',
                    facecolor = 'black',
                    alpha = 0.5))

# Palisades label (bottom left)
ax.text(0.15, 0.20,
        'Palisades Fire',
        transform = ax.transAxes,
        fontsize = 12,
        color = 'red',
        fontweight = 'bold',
        bbox = dict(boxstyle = 'round',
                    facecolor = 'white',
                    alpha = 0.75))

# ---------------------------------------------------------------
# TITLE AND LAYOUT
# ---------------------------------------------------------------

# Label the map to clarify the image and overlays
ax.set_title(("False Color Landsat (Feb 23, 2025) with Fire Perimeters (Jan 21, 2025)"),
             fontsize = 13)

# Remove both axis ticks
ax.set_xticks([])
ax.set_yticks([])

# Replace x and y label
ax.set_xlabel("Easting (meters)", fontsize = 12)
ax.set_ylabel("Northing (meters)", fontsize = 12)

# Improve layout so elements do not overlap
plt.tight_layout()

Figure Description

This map shows a false color Landsat composite with SWIR2, NIR, Red bands over the perimeters for both Eaton and Palisade fires. In this false color image, shortwave infrared (SWIR2) and near-infraed(NIR) energy responses highlight differences between healthy vegetation, burned, and developed areas. Burned area appear in different color compare to healthy vegetation areas are bright green. Overlaying the fire perimeter boundaries on top of the images shows the visual comparison of burn patterns.

Part 2 - Social Vulnerability

The 2024 Environmental Justice Index (EJI) dataset examined community characteristics in the census tracts that intersected each fire. All data were converted to a shared CRS and spatially joined to the fire perimeters using geopandas.sjoin().

To focus on social and demographic factors relevant to wildfire response, following indicator was selected: - E_AGE65: Percent population aged 65+

Setup for the Analysis

# Import library that is necessary for this analysis
import contextily as ctx

# EJI data (geodatabase)
eji = gpd.read_file("data/EJI_2024_California/EJI_2024_California.gdb")

# Reproject
eji = eji.to_crs(epsg = 4326)

Spatial Join - Identify Census Tract intersecting Each Fire

To determine which communities were affected, I performed a spatial join between the EJI census tracts and each fire perimeter. This identifies all tracts whose boundaries intersect the Eaton or Palisades fire footprint.

# Spatial join
census_within_palisades = gpd.sjoin(eji, palisades, how = "inner", predicate = "intersects")
census_within_eaton = gpd.sjoin(eji, eaton, how = "inner", predicate = "intersects")

# Check how many tracts there is
print("Palisades tracts:", len(census_within_palisades))
print("Eaton tracts:", len(census_within_eaton))

Palisades tracts: 36
Eaton tracts: 38

Polygon Clipping

To focus only on the census-tract areas actually affected by each fire, I clipped the EJI tracts to the Eaton and Palisades fire boundaries. This trims the polygons to the exact burned footprint and ensures the demographic analysis reflects only the portions of each tract that were inside the fire perimeter.

# Clip EJI census tracts to each fire perimeter
palisades_clipped = gpd.clip(census_within_palisades, palisades)
eaton_clipped = gpd.clip(census_within_eaton, eaton)

Plot Age 65+ for Both Fire Areas

After clipping the census tracts to each fire boundary, I mapped the percentage of residents aged 65 and older using a shared color range. This visualization illustrates how the elderly population is distributed within each burn area and highlights differences between the communities affected by the two fires.

# Assign name for the EJI column name
eji_variable = 'E_AGE65'

# Find common min/max for legend range
vmin = min(census_within_palisades[eji_variable].min(), 
           census_within_eaton[eji_variable].min())

vmax = max(census_within_palisades[eji_variable].max(), 
           census_within_eaton[eji_variable].max())

fig, (ax1, ax2) = plt.subplots(1, 2, figsize = (7, 7))

# Plot census tracts within Palisades perimeter
census_within_palisades.plot(
    column = eji_variable,
    vmin = vmin, vmax = vmax,
    legend = False,
    ax = ax1)
ax1.set_title('Palisades Fire Area', fontsize = 25)
ax1.axis('off')

# Plot census tracts within Eaton perimeter
census_within_eaton.plot(
    column = eji_variable,
    vmin = vmin, vmax = vmax,
    legend = False,
    ax = ax2)
ax2.set_title('Eaton Fire Area', fontsize = 25)
ax2.axis('off')

# Add overall title
fig.suptitle('Age 65+ - Fire Areas Comparison', fontsize = 45)

# Add shared colorbar at the bottom
sm = plt.cm.ScalarMappable(norm = plt.Normalize(vmin = vmin, vmax = vmax))
cbar_ax = fig.add_axes([0.25, 0.08, 0.5, 0.02])  # [left, bottom, width, height]
cbar = fig.colorbar(sm, cax = cbar_ax, orientation = 'horizontal')
cbar.set_label('Population Age Over 65')

plt.show()

Figure Description

Clipping the EJI census tracts to each fire boundary refines the analysis by showing only the precise areas affected by the Eaton and Palisades fires. When mapped, the clipped tracts reveal that each fire intersects a different set of communities. Using the Age 65+ indicator, the Palisades area generally shows a higher proportion of older adults, while the Eaton area displays lower but more variable elderly populations. This difference highlights how the fires impacted communities with distinct demographic vulnerabilities.

Reference

• Los Angeles GeoHub / NIFC FIRIS. (2025). Palisades–Eaton dissolved fire perimeters [data file]. Available: https://geohub.lacity.org/maps/ad51845ea5fb4eb483bc2a7c38b2370c/about. [Accessed: Nov. 15, 2025]

• U.S. Geological Survey. Landsat Collection 2 Level-2 Surface Reflectance (Microsoft Planetary Computer version) [data file]. Available: https://planetarycomputer.microsoft.com/dataset/landsat-c2-l2. [Accessed: Nov. 15, 2025]

• U.S. Department of Health and Human Services. (2024). Environmental Justice Index (EJI), 2024 – California Census Tract Data [data file]. Available: https://www.atsdr.cdc.gov/place-health/php/eji/eji-data-download.html. [Accessed: Nov. 21, 2025]

• Wikimedia Commons. (2025). 2025 Southern California Fires and the United States Forest Service (USFS) – Taskforce 1600 at the Palisades Fire (54264940069).jpg [image file]. Available: https://commons.wikimedia.org/wiki/File:2025_Southern_California_fires_and_the_United_States_Forest_Service_(USFS)_-Taskforce_1600_at_the_Palisades_Fire(54264940069).jpg. [Accessed: Dec. 4, 2025]